Skip to content

sklearn.py: expose number of CV folds and scoring metric from the config file....

Patrick Godwin requested to merge sklearn_crossval_defaults into master

Couple of changes to cross-validation within sklearn classifiers:

  • expose number of CV folds and scoring metric from the config file.
  • change the default number of CV folds from 3 to 5 to use more data at the expense of longer train times, is now configurable though.
  • change the default number of processes used in CV from 8 to 1 to explicitly force the user to think about using more cores. This should alleviate CPU/memory request issues when using the condor workflow.

Merge request reports